minimal clinically important difference

Zhang, Xi and Huang, in Health and Quality of Life Outcomes, trace the term to Jaeschke and Guyatt, who in 1989 defined the minimal clinically important difference as the smallest difference in score in the domain of interest which patients perceive as beneficial and which would mandate, in the absence of troublesome side effects and excessive cost, a change in the patient's management. The Cochrane Handbook puts the same idea more briefly, as the smallest change in instrument score that patients perceive is important, and says knowing it can greatly facilitate the interpretation of results. The point of it is to separate a result that is statistically distinguishable from zero from a result worth having.

There is no single way of arriving at the number. The review lists four in current use: the anchor-based method, which sets the target measure against an external one such as a patient's own rating of whether they improved; the distribution-based method, which takes a fraction of the spread in the data, commonly half a standard deviation; a literature analysis that synthesises published values; and expert consensus. The authors note that estimates reached from a statistical perspective alone, as the distribution-based method does, cannot scientifically explain the threshold they produce.

The thresholds that come out differ enough to matter. Looking at 28 studies that all calculated a threshold for the same instrument, the SF-36 quality of life questionnaire, the review found published values running from 2 to 17.4 on its physical component and from 1.46 to 10.28 on its mental component, with different anchors and different statistical methods giving different answers. The Cochrane Handbook adds a warning pointing the other way: a difference smaller than the threshold may be read as trivial when a substantial proportion of patients have in fact achieved an important benefit. So when a paper reports that a result did or did not reach clinical importance, the threshold it used is part of the claim, and where the authors set that threshold themselves rather than adopting an established one, the choice is worth reading.